By skyforbes If you only use image generation for one-off, single-shot images, you probably won't perceive my critique. But for those of us using continuous generation with 🍌Banana Pro — where you need to maintain continuous attributes across dozens of frames, keeping details in characters and scenery consistent — the drop in quality is undeniable.
🍌🍌🍌🍌🍌🍌🍌🍌🍌🍌🍌
The Launch (Day 1): The images were perfect. One prompt and you were done. No adjustments needed. It felt like a massive leap forward. Huge project and studio started in just 20 hours.
Day 2 & 3: As the user base likely spiked, the generation — on Pro — started getting "Dumber". To achieve a result with the same creative standard as Day 1, it suddenly required 5 to 8 adjustments and pure persistence 🙄.
Day 4 onwards: It became practically unusable. The system, in the Pro version, is showing signs of extreme stupidity 😴. It can no longer maintain logic for more than 2 prompts. It fails basic commands now, even with extremely refined and professional prompting. And drop from Pro 🍌 after 5 to 15 images a day.
I believe Google has found a "balance" between reducing the intelligence 🐵 of the model and drastically reducing 📉 usage limits for us, paid users 🤬.
It is unfortunate, but this seems to be the reality. People keep saying there is an "AI Bubble 🫧" on the way, but the demand is clearly proving to be infinitely superior to the current capacity. We are seeing this throttling even with heavy funding and subscriptions at a steep price point.